Scientific Explorations through Automatic Tone Recognition: From Speech Analysis to Model-Brain Alignment
- Speaker: Prof. YUAN Jiahong
(Professor
School of Humanities and Social Sciences, University of Science and Technology of China) - The rapid advancement of artificial intelligence in recent years has created new opportunities for research at the intersection of linguistics, neuroscience, and AI. Taking speech recognition as an example, the application of Transformer models and the pre-training-fine-tuning approach has significantly improved recognition accuracy, including recognizing tones in connected speech. Building on my research in automatic tone recognition, this talk presents a method for speech analysis using model representations and discusses three studies related to model-brain alignment: learning mechanisms and capacity limits of models, encoding processes of normalization, and neural decoding of tones.
- Date: 4 December 2024 (Wednesday)
